External and Internal Determinants of Development

Thomas Osang

s Rodrik, Subramanian, and Trebbi (2004) point out, factors that affect economic development can be classified using a two-tier approach. Based on a standard production function, inputs such as labor and physical and human capital directly affect per capita income. Much of the empirical cross-country growth literature has focused on these covariates. But the factors themselves are the product of deeper and more fundamental determinants and, thus, are at best proximate factors of economic development. The deeper determinants fall into two broad categories: internal and external. Among the former, institutions and geography have received the most attention, while international trade has been the focus of the latter.1 The main purpose of this paper is to add an external factor, namely measures of migration, to the existing geographyinstitutions-trade setup and to evaluate its contribution to the observed differences in per capita income across countries. Geography refers to the physical location of a nation and the various physical characteristics it is endowed with (for instance, distance from the equator, access to sea, agro-climatic zone, disease environment, soil type, and natural resources). A countrys size, access to sea, and general topography can crucially affect transport costs and the extent of its integration with the world. Climate and soil affect the types of crops planted. Interestingly, geography may even contribute to the nature of a countrys early institutions (Gallup, Sachs, and Mellinger 1998; Sachs 2003). Thus, geography is an obvious choice as an essential factor that shapes the course of a nations development. The role of institutions for development can be directly linked to the work of Douglass North (1993; 1994a, b, c). Norths motivation was the inability of neoclassical theory to explain widespread differences in economic performance
35

36

Thomas Osang

across countries. If only factor accumulation led to progress, all countries would advance, provided the payoffs are high enough. Since progress is absent in many countries, the payoffs must be different for different countries and institutions may be the reason for the differences (North 1994a). Institutions dene the rules of the game that determine the incentives people face and the choices they make. An alternative way of looking at institutions is through the transaction-cost approach. Well-dened rules and their smooth enforcement for example, better institutional quality greatly reduce transaction costs faced by economic agents and, thus, lead to more efcient economic outcomes (North 1993; 1994b). One of the rst studies to carefully examine the impact of institutions on productivity levels across countries was Hall and Jones (1999). Unlike geography, however, there is a potential endogeneity problem with institutions that needs to be addressed in the empirical investigation. International trade may affect economic development in several ways. In addition to gains from specialization in production based on comparative cost advantages, trade can make available new technologies and ideas, which, in turn, enhance total factor productivity. Moreover, operating in a larger market allows rms to take advantage of economies of scale and consumers to take advantage of a larger variety of goods. The empirical literature on the international tradedevelopment nexus is extensive, but a few papers stand out. Sachs and Warner (1995) construct an openness index and nd that greater openness leads to higher growth. Similarly, Frankel and Romer (1999) nd that international trade plays an important role in explaining cross-country differences in economic performance. Since trade measures, too, are likely to be endogenous, the authors construct an instrument for trade using a gravity-type model that explains the volume of trade between countries through their joint economic size and the distance between them. Migration can affect development in numerous ways, such as changes in the cost of labor, the loss or gain of human capital, knowledge spillovers, or workers remittances. While empirical literature on the impact of remittances is fairly extensive,2 fewer studies examine the role of emigration of skilled workers (brain drain) or the potential brain gain due to migration (Beine, Docquier, and Rapoport 2001). This study indicates that both internal and external determinants matter for development. The internal measures institutions and geography exhibit the expected signs and are typically statistically signicant, but they differ in their economic impact. Institutional measures appear to have large elasticity estimates, while geography measures are rather small. Among the external determinants, trade measures and the foreign-born population share (destination-country measure) exhibit the expected signs and are signicant in most specications. Interestingly, remittances (source-country measure) appear to contribute little to the

External and Internal Determinants of Development

37

observed variation in per capita income across countries unless the sample is restricted to the top half of all countries receiving remittances. In that case, remittances have a positive impact on economic development. In the next section, we provide an overview of the literature on trade, migration, and development. In the sections that follow, we describe the empirical models used, discuss the data set, and present and interpret the empirical results.

Review of the Literature

In this section, we review the literature on the trade, migration, and development nexus. We begin with the welfare and labor market implications of migration as well as a brief discussion of political economy issues related to migration. We then investigate the special relationship between trade and migration in the context of their joint effect on economic development. Finally, we review a number of papers that examine the impact of both migration and trade within a regional context. Migration and Development Welfare Effects of Immigration. From an empirical standpoint, there is no agreement on the gains or losses from immigration at the aggregate (national) level for either destination or source country. Martin (2003) maintains that economic gain from the current level of immigration in the United States is small, and even doubling the number of entering migrants would not make a great deal of difference. Head and Ries (1998) suggest in passing that immigration lowers transaction costs and generates trade gains that would not have been realized otherwise. In a welfare analysis, Razin and Sadka (1997) determine that those left behind in the source country lose, landlords in the destination country gain, and wage earners in the destination country lose, though their loss is less than the gain of the landlords. While the Razin and Sadka ndings like the majority of studies in the migration literature3 suggest a net gain in the destination country, an empirical study by Davis and Weinstein (2002) nds that U.S. natives collectively suffer a $72 billion loss per year due to migration, roughly equal to 0.8 percent of gross domestic product. Davis and Weinstein argue that immigration increases the output of the destination country while decreasing the output of the source country. The net effect in the U.S. is deterioration of trade as prices for U.S. goods go down while those of foreign goods go up. Furthermore, gains that accrue to the immigrants source country may be greater than the loss sustained by the destination country.

38

Thomas Osang

Labor Market Effects of Migration. A key result of the HeckscherOhlin model is the Rybczynski theorem, which states that a difference in a countrys endowment of labor will be reected in its output of goods. Gandal, Hanson, and Slaughter (2000) cast some doubt on the empirical validity of the theorem in a case study of Israeli immigration. In the early 1990s, Israel experienced a massive inux of highly skilled Russian immigrants (relative to the Israeli population). Curiously, this did not signicantly depress the wages of Israeli workers. The authors show that the mix of output in Israel did not change during this period to reect the change in labor composition. The most skill-intensive industries were not always the fastest growing. Instead, a global wave of skill-biased technological change helped Israel adjust to such a shock in factor supply. In fact, the change in production technology was such that the effective supply of skilled labor in Israel decreased even as its raw supply increased. The technological advances could have come to Israel from the United States through bilateral trade, capital ows, and government activities. Interestingly, in a related study using U.S. state-level data, Hanson and Slaughter (2002) nd evidence in support of the Rybczynski effects. Political Economy of Immigration. Rising differentials in global per capita income and advances in technology and transportation have contributed to an upsurge in international migration ows. Russell and Teitelbaum (1992) nd this increase is most dramatic among illegal migrants. Futhermore, migration movements have become not only greater but also more volatile and unpredictable and are accompanied by signicant remittance ows. They also play a role in the trade of many services previously considered nontradable. This trend raises concerns in wealthy countries, where the native populace is often resistant to immigration because of its potential to depress wages, displace native workers, or benet from wealth redistribution tax schemes. Dolmas and Huffman (2004) model the behavior of a voting population when it decides on the level of immigration. A critical determinant is the natives initial wealth level. Those endowed with relatively more capital will allow maximum immigration because the inux of migrants raises the marginal product of capital. Natives endowed with relatively less capital have to rely comparatively more on labor for their income, and since immigration erodes the marginal product of labor, poorer natives optimal decision is to allow zero migration. Interestingly, the natives collective decision is associated with the populations level of wealth inequality: Greater inequality is likely to lead to a no-immigration policy, while inequality that approaches zero can bring a maximum-immigration policy. In the survey article by Razin and Sadka (1997), the potential loss suffered by the native population in a welfare state through wealth redistribution tax policies is given as a possible reason for native resistance to immigration. The Dolmas and Huffman model also addresses this scenario and shows that the tax rate approaches zero as

External and Internal Determinants of Development

39

the number of voting immigrants approaches 100 percent of the original population. However, the tax rate rises signicantly once immigrants outnumber natives. Razin and Sadka also mention that the reallocation of investment from physical to human capital further erodes native welfare. The Relationship Between Trade and Migration Migration and Trade as Substitutes. If the fear is tenacious that immigration may result in losses for the host country, the HeckscherOhlin model suggests one approach for reducing the ow: The unimpeded movement of goods will lead to the equalization of factor prices, and that will remove an incentive for labor to move from one country to another. Horiba (2000) nds empirical evidence of the HeckscherOhlin theory. He shows that the convergence toward a more similar relative labor supply (which would equalize wages) is limited in magnitude, perhaps due to the costs associated with migration. Instead, the trade in goods, which can be considered trade in the factors that produced these goods, follows the same path as one would expect the factors to move according to HeckscherOhlin. Migration and Trade as Complements. Razin and Sadka (1997) point out that trade and immigration are substitutes only under the somewhat restrictive conditions of the HeckscherOhlin framework, allowing for country differences in the relative factor endowments only. For countries that differ in other aspects technology, for example free trade cannot equalize factor prices and may even widen factor price differentials. Immigration will allow each country to further specialize in the goods in which it has a technological advantage, leading to complementarity between trade and migration. Helliwell (1997) and Head and Ries (1998) both offer empirical evidence that trade and migration are complementary insofar as migration is capable of facilitating trade. Specically, Head and Ries nd that a 10 percent increase in immigration in Canada is associated with a 3 percent increase in imports and a 1 percent increase in exports to the immigrants source country. They attribute this nding to two factors: Immigrants may have a preference for goods produced at home, and immigrants knowledge about their home economies can lower the cost of foreign trade. However, the authors note that the tendency for immigration to increase imports more than exports creates a decrease in net exports, which can translate into currency depreciation and a loss of welfare for the destination country, though such a loss can be offset by social and economic gains that accrue from increased diversity. Migration and Trade as Complements in the Short Run and Substitutes in the Long Run. Two recent papers, both theoretical in nature, conjecture that the relationship between trade and migration depends on the time horizon of the analysis.4 Ludema and Wooton (1997) nd that trade liberalization is initially

40

Thomas Osang

agglomerative, creating a manufacturing core that attracts labor from the country of origin (an argument earlier posited by Krugman 1991), but as trade liberalization continues, the cost of trade becomes sufciently low that a manufacturing core loses its advantages and some labor shifts back to the periphery. Ludema and Wooton emphasize the importance of timing revealed by this diversicationagglomerationdiversication pattern and suggest that countries in the midst of liberalizing trade ought to restrict labor mobility until agglomerative forces weaken. Lopez and Schiff (1998) deconstruct migration patterns by skill composition in a small, labor-abundant developing economy after unilateral trade liberalization. Initial liberalization does not have much effect on the movement of skilled labor, but it does increase the number of unskilled workers leaving the country. The total labor force decreases, though the average skill level of the remaining population rises. Once trade has become substantially liberalized, the number of unskilled emigrants decreases and the total labor force stabilizes. This result is consistent with the pattern described above: a temporary spike in migration followed by a stabilization of migration ow. Immigration and Trade: Regional Analysis Taylor (1995) looks at the AsiaPacic region and, in particular, the determinants of the regions relatively high economic growth rates. His empirical ndings point to the high investment rate, primarily imported capital, as the biggest factor. Secondary causes include human capital accumulation and low population growth. Migration plays a very limited role, partly because the movement of people has become relatively restricted, unlike the massive immigrant ows that characterized the pre-World War I days. Taylor does suggest, however, that the movement of goods may have substituted, to some extent, for the movement of labor. Examining the effects of immigration and trade on a host countrys wage structure, Borjas et al. (1997) perform an empirical study on the U.S. labor pool and nd that neither immigration nor trade can be counted as a sufcient explanation for the widening differential between unskilled and skilled wages. However, in the case of native workers with less than a high school education, immigration has a decidedly large effect on the relative wages, more so than trade. The magnitude of this impact can be attributed to the ow of less-educated immigrants into the country, which raised the relative supply of unskilled workers (those without a high school education) 15 to 20 percent between 1980 and 1995. The authors do concede, however, that isolating the effects of immigration on the native labor market is difcult, in part because immigration does not have large regional effects. The movement of native migrants tends to balance that of immigrants so that relative skill endowments stay the same. As a result, comparing regions

External and Internal Determinants of Development

41

especially receptive of immigrants to other regions does not provide meaningful results. In addition, other factors that inuence the U.S. labor market are not adequately controlled, and a realistic counterfactual is difcult to establish. Dunlevy and Hutchinson (1999) base their paper upon recent ndings that immigrants have a pro-trade effect between source and host country (Head and Ries 1998). Data on U.S. trade and immigration between 1870 and 1910 provide empirical evidence for this pro-trade effect, particularly on nished foodstuffs and manufactures. For these two categories of goods, a 10 percent increase in migrant stock increased imports from the source country by 4 percent. The authors also nd that the pro-trade effect diminished or was nonexistent for New European countries (eastern and southern Europe) as well as for the period between 1900 and 1910. They hypothesize that immigrants from New Europe were unable to form the kind of links or relationships that would facilitate trade. They also suggest that from 1900 to 1910, a signicant shift in source countries occurred, which also weakened the pro-trade effect. In general, this study supports the Head and Ries paper, which focuses exclusively on U.S.Canada trade. Martin (2003) also focuses on post-NAFTA Mexico, though his paper is more descriptive than empirical. He argues that when the assumptions involved in HeckscherOhlin are relaxed, trade and migration are more likely to be complements. He calculates that migration to the United States will increase by 10 to 30 percent in the ve to fteen years following the North American Free Trade Agreement, creating what he terms a migration hump. He does predict, however, that migration will decrease soon after due to social and economic trends in Mexico. Robertson (2005) takes an empirical approach to the Mexican labor market following NAFTA. He uses wage convergence as a measure of labor market integration and nds that the rate of wage convergence in post-NAFTA Mexico did not signicantly increase. Integration was not uniform, as one would expect if trade were the main force behind wage convergence, nor was it higher in manufacturing industries that received large amounts of foreign direct investment. Instead, integration was highest in the two border citiesTijuana and Ciudad Jurezthat experienced large immigrant ows. Robertson thus concludes that migration plays the most signicant role in labor market integration. Liberalization in trade and capital ows alone is insufcient to induce wage equalization. This, of course, contradicts the HeckscherOhlin premise of factor price equalization following free trade.

Empirical Models
The starting point of our empirical investigation into the internal and external determinants of economic development is the following linear empirical model:

42

Thomas Osang

(1)

Incomei= 1+ 2Insti+ 3Geogi+ 4Tradei+ 5Migi+ ei

where Incomei is income per capita in country i and Insti, Geogi, Tradei, and Migi are measures of country is institutions, geography, international trade volume or policy, and migration, respectively. As mentioned before, institutions and geography represent the internal determinants of development, while trade and migration are the external measures. Simple least square (OLS) estimates of equation (1) will serve as the benchmark for subsequent specications. Our second empirical specication addresses the issue of endogeneity of regressors. Institution, trade, and migration measures are likely to be endogenous due to measurement error, survey bias, and/or reverse causality.5 Consequently, appropriate instruments are needed for all measures. Of the various external instruments found in the literature, two stand out due to their widespread use: settler mortality as an instrument for institutions (Acemoglu, Johnson, and Robinson 2001),6 and predicted trade shares as an instrument for a countrys actual trade share (Frankel and Romer 1999) (Table 1). Since the exogeneity of the geography measure is indisputable (and assuming for now that migration is exogenous), our second specication is the two-stage least square (2SLS) estimator version of equation (1), with the following rst-stage regressions for the two endogenous regressors (institutions and trade): (2a) (2b) Insti = a1+ a2SMi+ a3PTradei + a4Geogi + 5Migi + i Tradei = b1+ b2SMi+ b3PTradei + b4Geogi + b5Migi + i

where SMi measures settler mortality and PTradei is the predicted trade share in country i. One problem with specications (2a) and (2b) is that the two instruments used are highly correlated. As a result, they may not be able to identify the impact of the endogenous regressors they are instrumenting for (Dollar and Kraay 2003). As an alternative to the external instruments, our third specication uses internal instruments instead (Lewbel 1997). In particular, we use second- and third-order-central moments of the endogenous variables as instruments.7 In this specication, we not only account for the potential endogeneity of the trade and institution measures but of the migration measures as well. Finally, a shortcoming of all the above models is that they assume that all covariates have the same impact for all countries. In other words, the model ignores unobserved time-invariant heterogeneity across countries. Using a paneldata approach enables us to exploit the time dimension of the data to account

External and Internal Determinants of Development

43

Table 1

Variable Definitions and Data Sources Institutions

Name CIM (contract intensive money) Definition and source(s) Defined as the ratio of noncurrency (M1 minus currency) to total money (M2). Compiled by R. M. Bittick, California State University, Dominguez Hills, based on data from the International Monetary Fund (1998). Measures the quality of contract enforcement, police and courts, as well as the likelihood of crime and violence, average for 1996, 1998, and 2000. From Kaufmann et al. (2003)

Rule of law

Trade
Name Trade share Import tariffs Definition and source(s) Imports plus exports relative to GDP. From PWT Mark 6.1 (Heston et al. 2002). Import duties as a percentage of total imports. From World Bank (2003), authors calculations.

Geography
Name Distance equator (relative distance from the equator) Name Remittances share Foreign population share Definition and source(s) Calculated as distance from the equator, divided by 90. From Gallup et al. (1998) and Hall and Jones (1999).

Migration
Definition and source(s) Ratio of remittances to GNP. From World Bank (2000). Ratio of foreign born to total population. From United Nations (1994). Definition and source(s) Mortality rate of European colonialists in the 1500s. From Acemoglu et al. (2001). Obtained from bilateral gravity-type equations and controlling for geography. From Frankel and Romer (1999).

Instrumental variables (external) for 2SLS regressions

Name Settler mortality Predicted trade share

44

Thomas Osang

for this unobserved country-specic heterogeneity. The following panel model is estimated: (3) Incomeit= 1+ it+ 2Instit+ 3Geogit+ 4Tradeit+ 5Migit+ it

where i and t are country- and time-specic xed effects, respectively. A paneldata specication such as (3), however, is problematic if some of the right-handside variables are time invariant. Thus, when using mean- or rst-differencing to remove the unobserved time-invariant country-specic effects the standard procedure in xed effect (FE) estimation all time-invariant covariates such as most geography measures are removed from the estimation equation as well. However, the lost parameter estimates can be recovered through an auxiliary regression of the estimated xed effects on the time-invariant covariates. As an alternative to the above FE model, we also estimate a random effects (RE) model. Hausman specication tests guide us in the model-selection procedure.

Data
In general, the data set covers the four decades from 1961 to 2000, though fewer time periods may be available for certain variables. For the cross-section estimates, all time-varying variables are averaged, except for the dependent variable that is measured in 2000. The number of countries varies among the different specications of the baseline regression model, ranging from N = 65 to N = 125. For the panel-data estimates, the time-varying variables are averaged over 10 years to smooth out temporary shocks and business cycle uctuations common across countries. As a result, the time dimension of the panel-data regressions includes four years of observations. The number of countries in the panel regressions is N = 68. Dependent Variable. Our measure of economic development (the dependent variable in all regressions) is the log of per capita GDP in 2000, expressed in purchasing power parity-adjusted dollars (Heston, Summers, and Aten 2002). Explanatory Variables. Our main measure of institution is contract intensive money (CIM), which was proposed by Clague et al. (1999). It is dened as the ratio of noncurrency money to total money. The basic argument for such a measure stems from the fact that in societies where the rules of the game and property and contract rights are well dened, even transactions that heavily rely on outside enforcement can be advantageous. Currency in this setting is used only in small transactions. Agents are increasingly able to invest their money in nancial intermediaries and exploit several economic gains. Clague et al. discuss the various gains from increased use of CIM and augment their use of CIM with

External and Internal Determinants of Development

45

case studies. They also show that CIM is a measure of contracting environment and not of nancial development, as one might suspect. This measure is thus in line with the denition of institutions as noted above. Moreover, CIM is a rather objective measure without the many biases and measurement errors that are typical of the survey-based measures of institutions. While CIM is our preferred measure of institutions, we also use the rule of law (Kaufmann, Kraay, and Mastruzzi 2003) as an alternative measure. This variable captures the extent to which agents abide by the rules of society. To control for the effect of geographic location and climate, we use a countrys distance from the equator (distance equator). We measure the extent of a countrys openness to international commerce in two ways: by its trade share, as dened by the ratio of exports and imports to GDP, and by the average import tariff, constructed as the ratio of import duties to imports (tariff rate). A countrys exposure to international migration is also captured in two ways. First, we use the share of remittances in gross national product (remittances share) as an indicator of the potential benets from emigration for the source country. Second, we employ the ratio of foreign born to total population (foreign-born share). The foreign-born share can be interpreted as measure of the potentially benecial impact of immigration for the destination country, either as a proxy for the size of the immigration surplus or the positive externalities associated with immigration. A third migration measure, the ratio of emigrants to total native population, can be interpreted as an indicator of the negative brain-drain effect of emigration for the source country. However, reliable emigration data are either difcult to obtain or not available. For this reason, we do not consider the measure in this study.8

Empirical Results
Cross-Section Estimates To contrast our empirical results with the literature, we rst estimate the crosssection specication used by Rodrik, Subramanian, and Trebbi (2004); see Table 2A, col. 1. The measure of economic development is the log of per capita GDP, expressed in international prices.9 Openness to international trade is measured as the average trade share from 1961 to 2000. Rule of law is used to measure the quality of public institutions, while the measure of geography is distance from the equator, distance equator. While the magnitude of the coefcient estimates in col. 1 are not exactly identical to the ones reported in Rodrik, Subramanian, and Trebbi, all signs are the same and both geography and institutions are statistically signicant (at the 1 percent level), while trade share is not statistically signicant.

46

Thomas Osang

Table 2A

Cross-Section Regressions I: Remittances, Rule of Law, and Trade Share

(Dependent variable natural log GDP per capita in 2000, in purchasing power parity dollars) (1) (2) Model OLS OLS Rule of law .812 (11.7)** Trade share .0277 (.33) Distance equator 1.348 (4.05)** Remittances share 1.024 (7.14)** .0475 (.43) .736 (1.68) .0226 (.96) (3) 2SLS IV Set A 2.849 (3.32)** .434 (1.00) 1.940 (.93) .131 (1.51) 42 .55 [.0008] (4) 2SLS IV Set B 2.735 (3.43)** .300 (.74) 1.342 (.69) .190 (1.93) 42 .41 [.0023]

Observations 131 80 R-squared .70 .48 DWH test: OLS (null) vs. 2SLS (P-value)

NOTES: t statistics in parentheses, **p<0.01, *p<0.05, p<0.1; P-values in square brackets; IV Set A: Settler mortality for rule of law; predicted trade share for trade share; IV Set B: Same as Set A plus higher-centered moments of remittances share. SOURCE: Authors calculations.

Adding remittances as a covariate (col. 2) does not lead to substantive changes in the results. This is not surprising given that the (negative) coefcient on remittances is insignicant by itself. However, the potential endogeneity of institutions, trade, and remittances has not been taken into consideration so far and, thus, all results may be biased. In column 3, we use the 2SLS estimator with settler mortality and predicted trade share as instruments while continuing to assume that the remittances share is exogenous. As in Rodrik, Subramanian, and Trebbi, using instruments for institutions and trade makes the geography measure insignicant without changing the lack of signicance of the trade measure and the remittances share. When we use instrumental variables to account for the potential endogeneity of the remittances share (col. 4), both institutions and remittances are signicant at 1 percent and 5 percent, respectively, while geography and trade measures remain insignicant. Given the negative sign on the remittances share, remittances appear to have a negative impact on a countrys macroeconomic performance. At this point, our results appear to support Rodrik et al.s conclusion that institutions rule, with the added twist of the negative impact of remittances. Note that the DurbinWuHausman (DWH) test easily rejects the simple OLS models (col. 2) in favor of the two 2SLS estimators (cols. 3 and 4).

External and Internal Determinants of Development

47

Next, we reestimate Table 2A with different and, we believe, better measures of institution and trade. Instead of rule of law, we use CIM for institutions, while trade share is replaced with a trade policy measure, the average import tariff. In addition, we use internal instruments for trade and institutions rather than the external instrumental variables in Table 2A. The results are shown in Table 2B. The immediate consequence of the substitutions is a reduction in the sample size for the OLS estimates (cols. 5 and 6), while the 2SLS sample sizes are slightly larger (cols. 7 and 8). In terms of the estimates, the main differences pertain to the impact of geography and trade. In every specication of Table 2B, trade and geography measures have the expected signs and are signicant at least at the 10 percent level. The remittances share, however, is negative and insignicant throughout. Note that at the 10 percent level, the DWH test indicates that OLS (col. 6) is preferred over the 2SLS estimates. So far, the skewed nature of the remittances variable has not been taken into consideration. As Table 3 reveals, more than half the countries listed have remittance shares that are less than 1 percent of GNP. To account for this, we construct two remittance dummies, one for countries with shares between 1 percent and 10 percent (medium remittances share), and one for countries with shares larger
Table 2B

Cross-Section Regressions II: Remittances, CIM, and Tariff Rate

(Dependent variable natural log GDP per capita in 2000, in purchasing power parity dollars) (5) (6) Model OLS OLS CIM 2.529 (6.73)** Tariff rate .152 (2.56)* Distance equator 2.602 (5.45)** Remittances share 2.438 (5.03)** .345 (1.73) 1.473 (2.02)* .0263 (.57) (7) 2SLS IV Set A 2.029 (3.58)** .453 (2.23)* 1.560 (2.23)* .0160 (.36) 47 .52 [.3173] (8) 2SLS IV Set B 2.201 (3.80)** .405 (1.94) 1.671 (2.37)* .0616 (1.12) 47 .51 [.1492]

Observations 86 47 R-squared .73 .53 DWH test: OLS (null) vs. 2SLS (P-value)

NOTES: t statistics in parentheses, **p<0.01, *p<0.05, p<0.1; P-values in square brackets; IV Set A: Higher-centered moments of CIM and tariff rate; IV Set B: Higher-centered moments of CIM, tariff rate, and remittances share. SOURCE: Authors calculations.

48

Thomas Osang

than 10 percent (high remittances share). The results are shown in Table 2C. Once again, we compare simple OLS (col. 9) with 2SLS (col. 10), in which we use high-order-central moments of CIM and tariff rate as instruments. The results indicate that discretization of the remittances variable does not change the outcome. The medium- and high-share dummies arent signicant, and both are negative. In our nal cross-section model, we drop from our sample all countries with a remittances share of less than 1 percent of GDP since remittances are likely to play no role in these countries. The results are given in Table 2D. In addition to the OLS estimates, we report 2SLS estimates using internal instruments for institutions and trade (col. 13) and institutions, trade, and remittances (col. 14). We also use the trade share instead of the trade policy measure. Despite the reduction in power (N = 24 in cols. 1214), the coefcient estimates for institutions, geography, and trade are statistically signicant (except for trade in col. 11) and have the expected sign, while the coefcient estimate on the remittances share is now positive throughout and even statistically signicant at the 10 percent level in two specications (cols. 12 and 13). The insignicance of the coefcient estimate on
Table 2C

Cross-Section Regressions III: Remittances Group

(Dependent variable natural log GDP per capita in 2000, in purchasing power parity dollars) (9) Model OLS CIM Tariff rate Distance equator Medium remittances share High remittances share 2.102 (4.86)** .306 (1.58) 2.417 (3.42)** .238 (1.25) .0920 (.18) (10) 2SLS IV Set A 1.773 (3.48)** .391 (2.01)* 2.503 (3.67)** .228 (1.25) .0830 (.17) 64 .49 [.3606]

Observations 64 R-squared .50 DWH test: OLS (null) vs. 2SLS (P-value)

NOTES: t statistics in parentheses, **p<0.01, *p<0.05, p<0.1; P-values in square brackets; IV Set A: Highercentered moments of CIM and tariff rate. SOURCE: Authors calculations.

External and Internal Determinants of Development

49

Table 2D

Cross-Section Regressions IV: High Remittances Sample

(Dependent variable natural log GDP per capita in 2000, in purchasing power parity dollars) (11) (12) Model OLS OLS CIM 3.85 (7.42)** Trade share .0003 (.00) Distance equator 3.25 (6.86)** Remittances share 2.16 (3.89)** .356 (2.08)* 3.21 (4.26)** .17 (1.72) (13) 2SLS IV Set A 2.351** (4.60) .362* (2.05) 3.183** (4.58) .166 (1.86) (14) 2SLS IV Set B 2.296** (4.56) .361* (2.05) 3.223** (4.63) .143 (1.48)

Observations 48 25 25 25 R-squared .83 .68 .73 .73 DWH test: OLS (null) vs. 2SLS (P-value) [.2279] [.4877]
NOTES: Sample is restricted to countries in which remittances share of GDP exceeds 1 percent; t statistics in parentheses, **p<0.01, *p<0.05, p<0.1; P-values in square brackets; IV Set A: Higher-centered moments of CIM and trade share; IV Set B: Higher-centered moments of CIM, trade share, and remittances share. SOURCE: Authors calculations.

remittances in the last specication (col. 14) is mitigated by the fact that the DWH test cannot reject the OLS null hypothesis (col. 12). Thus, it appears that for the group of countries with substantial unilateral foreign transfers, remittances appear to matter for a source countrys economic development, in addition to the effects of trade, institutions, and geography. Panel Estimations We use the panel-data approach to investigate the impact of the share of the foreign-born population on economic development. As previously discussed, the foreign-born share measures the impact of migration on destination countries, compared with the remittances share, which affects only source countries. Furthermore, while the remittances share is important for developing countries only, nontrivial foreign-born population shares can be found in both developing and developed countries (see Table 4 for a ranking of countries by the foreignborn share).10 Our estimation results are given in Table 5. Column 1 contains the random-effects specication since the Breusch-Pagan (BP) test reveals that RE is strongly preferred over pooled OLS (the null hypothesis). While institutions,
(Continued on page 53 )

50

Thomas Osang

Table 3

Ranking of Countries by Remittances-to-GNP Ratio

Rank Country 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 21 22 23 24 26 27 28 29 30 31 32 33 34 35 38 39 40 Lebanon Yemen, Rep. Samoa Eritrea Tonga Jordan Cape Verde Albania Egypt, Arab Rep. Dominica Morocco St. Kitts and Nevis Burkina Faso El Salvador Jamaica Haiti Pakistan Benin Sri Lanka Belize Vanuatu Comoros Tunisia Dominican Republic St. Vincent and the Grenadines Mali Grenada St. Lucia Sudan Croatia Turkey Macedonia, FYR Senegal Bangladesh Somalia Nicaragua Nepal Algeria Togo India Mean 197098 .312 .263 .245 .196 .172 .170 .169 .131 .075 .072 .067 .065 .064 .057 .051 .048 .046 .045 .045 .045 .044 .042 .041 .039 .039 .038 .035 .031 .027 .025 .023 .021 .020 .019 .017 .017 .017 .014 .013 .011 Rank Country Honduras 42 Nigeria 43 Guatemala 44 Barbados Mexico 46 Seychelles Philippines Colombia Mauritania 50 Armenia Oman 52 Cameroon Niger Ecuador Cambodia 56 Peru RanCountDjibouti Mongolia Guinea-Bissau 60 Costa Rica Panama Indonesia Guinea Trinidad and Tobago Poland 66 Madagascar Sao Tome and Principe Belarus Ghana China Korea, Rep. Rwanda Moldova Brazil Paraguay Guyana Kyrgyz Republic Bolivia Congo, Dem. Rep. Mean 197098 .011 .010 .009 .008 .008 .006 .006 .006 .006 .005 .005 .004 .004 .004 .004 .003 .003 .003 .003 .002 .002 .002 .002 .002 .002 .001 .001 .001 .001 .001 .001 .001 .001 .001 .001 .001 .001 .001 .001

NOTE: Countries with no remittances were omitted. SOURCE: World Bank (2000); authors calculations.

External and Internal Determinants of Development

51

Table 4

Ranking of Countries by Share of Foreign-Born Population

Rank Country 1965 1975 1985 1990 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 22 23 24 25 26 28 29 30 34 35 37 38 39 41 42 44 United Arab Emirates Kuwait Qatar Macau Hong Kong Israel Jordan Bahrain Luxembourg Oman Singapore Australia Cote DIvoire Saudi Arabia New Zealand Switzerland Canada Taiwan Gambia France Libya Lebanon Belgium Argentina Malawi Sweden Pakistan Zimbabwe Somalia Ireland Puerto Rico Iran Syria United States Malaysia Gabon Zambia Togo Venezuela Costa Rica United Kingdom Burundi Uganda Congo Poland .842 .529 .515 .674 .466 .560 .362 .202 .158 .080 .282 .179 .024 .065 .146 .129 .017 .171 .118 .090 .043 .080 .055 .110 .075 .050 .106 .050 .004 .032 .023 .091 .061 .050 .088 .040 .092 .087 .063 .023 .045 .048 .109 .044 .070 .675 .524 .588 .601 .430 .416 .261 .213 .192 .158 .235 .195 .218 .115 .158 .162 .152 .114 .100 .105 .098 .076 .078 .083 .056 .069 .054 .060 .096 .053 .043 .044 .060 .054 .072 .040 .063 .062 .057 .014 .056 .038 .077 .047 .053 .663 .594 .586 .536 .405 .339 .261 .329 .270 .319 .186 .219 .261 .285 .151 .140 .151 .085 .094 .108 .143 .104 .090 .060 .040 .078 .065 .078 .093 .080 .099 .059 .068 .070 .047 .078 .045 .047 .061 .043 .062 .070 .013 .052 .040 Mean 196590

.900 .770 .715 .590 .633 .580 .446 .564 .400 .425 .309 .406 .265 .287 .351 .274 .316 .234 .335 .223 .155 .215 .234 .206 .292 .199 .257 .181 .155 .152 .160 .148 .155 .118 .079 .112 .112 .106 .104 .102 .123 .102 .122 .096 .090 .078 .052 .076 .121 .073 .089 .071 .061 .071 .079 .067 .072 .066 .093 .064 .092 .064 .062 .064 .066 .064 .079 .063 .042 .062 .088 .062 .041 .060 .041 .059 .053 .058 .153 .058 .065 .057 .061 .055 .019 .055 .059 .050 .036 .050 (Continued on page 52 )

52

Thomas Osang

Table 4 (continued)

Ranking of Countries by Share of Foreign-Born Population

Rank Country 1965 1975 1985 1990 46 47 48 50 51 53 54 57 58 59 61 62 63 64 65 67 69 71 72 75 78 83 85 88 89 Netherlands Zaire Nigeria Liberia Paraguay Sierra Leone South Africa Austria Tanzania Norway Senegal Denmark Sudan Honduras Cameroon Ghana Burkina Faso Nepal Turkey Italy Greece Iraq Mali Guatemala Portugal India Algeria Yugoslavia Korea Spain Kenya Romania Chile Niger Brazil Thailand Bangladesh Japan Mexico Ethiopia Czechoslovakia Egypt Colombia Myanmar .029 .052 .002 .031 .028 .024 .048 .017 .041 .020 .047 .021 .019 .022 .034 .058 .008 .033 .029 .016 .008 .003 .023 .011 .007 .019 .016 .008 .005 .009 .017 .018 .012 .004 .009 .014 .009 .006 .005 .001 .004 .007 .004 .002 .025 .062 .161 .039 .037 .032 .038 .028 .040 .027 .037 .026 .019 .014 .029 .029 .017 .023 .022 .018 .013 .009 .024 .006 .017 .015 .014 .010 .008 .009 .012 .011 .011 .003 .011 .010 .010 .006 .004 .002 .004 .005 .004 .001 .054 .031 .003 .046 .048 .044 .029 .036 .025 .037 .017 .035 .049 .020 .024 .014 .034 .017 .019 .023 .030 .033 .014 .014 .021 .012 .009 .017 .017 .010 .008 .007 .007 .015 .009 .007 .008 .006 .006 .003 .005 .004 .003 .002 .078 .028 .003 .050 .043 .050 .030 .059 .023 .044 .025 .041 .033 .056 .024 .009 .047 .021 .020 .027 .032 .028 .012 .029 .014 .010 .015 .017 .021 .018 .007 .006 .008 .015 .008 .006 .007 .007 .008 .016 .006 .003 .003 .002 Mean 196590 .047 .043 .042 .042 .039 .037 .037 .035 .032 .032 .032 .031 .030 .028 .028 .027 .026 .024 .023 .021 .021 .018 .018 .015 .015 .014 .013 .013 .013 .011 .011 .011 .009 .009 .009 .009 .009 .006 .006 .005 .005 .005 .004 .002

NOTE: List shows countries reporting these data to the U.N. SOURCE: United Nations (1994).

External and Internal Determinants of Development

53

Table 5

Panel Regressions: Share of Foreign-Born Population

(Dependent variable natural log GDP per capita in 2000, in purchasing power parity dollars) (1) (2) Estimation method RE FE CIM Tariff rate Distance equator Foreign-born share 1.377 .982 (6.43)** (4.50)** .0706 .0684 (3.60)** (3.52)** 3.127 (7.40)** .100 .0257 (2.09)* (.48) (3) RE-IV Set A .755 (2.17)* .0653 (2.85)** 3.419 (7.62)** .111 (2.23)* (4) RE-IV Set B .814 (2.36)* .0624 (2.73)** 3.404 (7.61)** .117 (1.87) 175 68

Observations 175 175 175 Number of countries 68 68 68 R-squared .33 BP test: pooled (null) vs. RE (P-value) [.0000] Hausman test: RE (null) vs. FE (P-value) [.0000] Hausman test: RE (null) vs. IV RE (P-value) [.1613]

[.1783]

NOTES: z statistics in parentheses, **p<0.01, *p<0.05, p<0.1; P-values in square brackets; RE-IV Set A: Higher-centered moments of CIM and tariff rate; RE-IV Set B: Higher-centered moments of CIM, tariff rate, and foreign-born share. SOURCE: Authors calculations.

trade, and geography coefcients are signicant and have the expected sign, the most important nding pertains to the foreign-born share, which is positive and signicant. Interestingly, the RE result on the foreign-born share is not robust when we estimate an FE model instead of the RE model (col. 2). While institutions and trade continue to be signicant (and have the expected signs), the foreign-born coefcient estimate is now much smaller in size and insignicant.11 However, given the small time-series dimension (T = 4), the FE model suffers from overtting and the corresponding decline in degrees of freedom.12 In the next two columns, we thus revert to the FE specication but use different sets of instruments to account for the potential endogeneity of the institutions, trade, and foreign-born share. Initially (col. 3), we use internal instruments only for the institution and trade variables, while in the nal specication (col. 4), we add internal instruments for the foreign-born share. The random effects/instrumental variables (RE-IV) estimates

54

Thomas Osang

conrm the nding from the simple RE model (col. 1), namely that the foreignborn share has a positive impact on macroeconomic performance. Overall, the differences in the estimated coefcients between the RE and RE-IV models are small, with a small preference for the RE model (according to the Hausman test, we cannot reject the RE model at the 10 percent level).

Conclusions
To our knowledge, this paper represents the rst attempt to integrate measures of migration into a framework that analyzes the economic impact of the so-called deep determinants of development: institution, geography, and international trade. Using both cross-section and panel-data estimation methods, we nd that both measures of migration used in this studyremittances as a share of GNP (top half of receiving countries) and foreign-born relative to the total populationhave a positive impact on economic development even after controlling for institutions, geography, and trade. The ndings go beyond establishing correlations. Using instrumental variable methods to counter the estimation bias of the three potentially endogenous covariates (institutions, trade, and migration), the ndings provide evidence for a causal link between external measures (migration and trade) and per capita income. In terms of their economic impact, institutions appear to matter the most. This is especially true if we measure the quality of institutions by the extent of contract-intensive money in the economy, which has a high elasticity with respect to per capita income. Openness to trade (when measured as the average import tariff) and migration (when measured as the foreign-born share) exhibit point elasticities that are more than ten times smaller than the one for institutions. The (positive) economic impact of geography appears to be rather small. This study can be extended in several directions. First, better measures of migration for both source and destination countries are desirable. As mentioned before, the migration measures used here cannot account for the negative brain drain or the potentially positive brain gain of migration in countries with liberal emigration rules. Similarly, the destination-country measure (foreign-born share) cannot differentiate between the positive and negative externalities associated with immigration. In addition to the average effect captured in this study, identifying the size of both positive and negative migration effects for the destination country would be desirable from a policymaking perspective. Second, methodologies could be improved. In addition to the extremes of random- and xed-effects estimations, a middle-ground panel-data estimator such as HausmanTaylor (1981) could prove to be a superior specication and, thus, may produce lessbiased estimates.

External and Internal Determinants of Development

55

Notes
I would like to thank Jeffry Jacob, Nancy Yang, and Kaycee Washington for their excellent research assistance. Financial support through a research grant from the SMU University Research Council is gratefully acknowledged. 1 See Easterly and Levine (2003) for a detailed overview of deep-determinant literature. 2 See Gosh (2006) for a survey of the literature on remittances and development. 3 See, for example, the study by the National Research Council (Smith and Edmonston 1997) for an assessment of the overall impact of immigration on the U.S. economy. 4 Martin (2003) points to a similar conclusion, though his work is narrower in scope, focusing exclusively on Mexico following NAFTA. 5 See Frankel and Romer (1999) for a more detailed discussion of this issue. 6 Acemoglu, Johnson, and Robinson (2001) argue that settler mortality is a suitable instrument for institutions since it is not correlated with current income other than through current institutions. Settler mortality determined the colonization strategies, which, in turn, shaped past institutions and subsequently formed current institutions. 7 The second central moment about the mean is commonly known at the variance of a random variable, while the third central moment is related to the skewness of a random variable; that is, the degree to which the mass of the distribution is concentrated on the left or right side of probability distribution. 8 Furthermore, emigration ratio measures are unable to distinguish between the negative braindrain and positive brain-gain effect, with the latter a result of increased human capital accumulation in emigrant countries (Stark, Helmenstein, and Prskawetz 1998). 9 Note that in this study, income is measured in 2000, compared with 1995 in Rodrik, Subramanian, and Trebbi (2004). 10 Note that while the paper discusses migration flows throughout, here we use migration stocks as the explanatory variable due to the nonavailability of data on migration flows for all but a small number of countries and time periods. 11 As a time-invariant measure, distance from the equator is dropped from the equation in any FE specification. 12 Note that the Hausman test rejects the RE null in favor of the FE model specification.

References
Acemoglu, Daron, Simon Johnson, and James A. Robinson (2001), The Colonial Origins of Comparative Development: An Empirical Investigation,American Economic Review 91: 1369 1401. Beine, Michel, Frederic Docquier, and Hillel Rapoport (2001), Brain Drain and Economic Growth: Theory and Evidence, Journal of Development Economics 64: 275 89. Borjas, G. J., R. B. Freeman, L. F. Katz, J. DiNardo, and J. M. Abowd (1997), How Much Do Immigration and Trade Affect Labor Market Outcomes? Brookings Papers on Economic Activity 1: 1 90.

56

Thomas Osang

Clague, Christopher, Philip Keefer, Stephen Knack, and Mancur Olson (1999), Contract-Intensive Money: Contract Enforcement, Property Rights, and Economic Performance, Journal of Economic Growth 4 (June): 185 211. Davis, D. R., and Weinstein, D. E. (2002), Technological Superiority and the Losses from Migration, NBER, Working Paper no. 8971 (Cambridge, Mass.: National Bureau of Economic Research, June). Dollar, David, and Aart Kraay (2003), Institutions, Trade, and Growth: Revisiting the Evidence, World Bank Policy Research Working Paper Series, no. 3004 (Washington, D.C.: World Bank, March). Dolmas, James, and Gregory W. Huffman (2004), On the Political Economy of Immigration and Income Redistribution, International Economic Review 45 (November): 1129 68. Dunlevy, J. A., and W. K. Hutchinson (1999), The Impact of Immigration on American Import Trade in the Late Nineteenth and Early Twentieth Centuries, Journal of Economic History 59 (December): 1043 62. Easterly, William, and Ross Levine (2003), Tropics, Germs, and Crops: How Endowments Influence Economic Development, Journal of Monetary Economics 50 (January): 3 39. Frankel, Jeffrey A., and David Romer (1999), Does Trade Cause Growth? American Economic Review 89 (June): 379 99. Gallup, John Luke, Jeffrey D. Sachs, and Andrew D. Mellinger (1998), Geography and Economic Development, NBER Working Paper no. 6849 (Cambridge, Mass.: National Bureau of Economic Research, December). Gandal, N., G. H. Hanson, and M. J. Slaughter (2000), Technology, Trade, and Adjustment to Immigration in Israel, NBER Working Paper no. 7962 (Cambridge, Mass.: National Bureau of Economic Research, October). Gosh, Bimal (2006), Myths, Rhetoric, and Realities: Migrants Remittances and Development (Report to the International Organization for Migration, Brussels, September). Hall, Robert E., and Charles I. Jones (1999), Why Do Some Countries Produce So Much More Output Per Worker Than Others? Quarterly Journal of Economics 114 (February): 83 116. Hanson, G. H., and M. J. Slaughter (2002), Labor-Market Adjustment in Open Economies: Evidence from U.S. States, Journal of International Economics 57 (June): 3 29.

External and Internal Determinants of Development

57

Hausman, Jerry A., and William E. Taylor (1981), Panel Data and Unobservable Individual Effects, Econometrica 49 (November): 1377 98. Head, K., and J. Ries (1998), Immigration and Trade Creation: Econometric Evidence from Canada, Canadian Journal of Economics 31 (February): 47 62. Helliwell, J. F. (1997), National Borders, Trade and Migration, NBER Working Paper no. 6027 (Cambridge, Mass: National Bureau of Economic Research, March). Heston, Alan, Robert Summers, and Bettina Aten (2002), Penn World Table Version 6.1, Center for International Comparisons at the University of Pennsylvania (October). Horiba, Y. (2000), U.S. Interregional Migration and Trade (Paper presented at the Murphy Institute Conference on the Political Economy of Migration, Murphy Institute of Political Economy, Tulane University, March 24 25). International Monetary Fund (1998), International Financial Statistics Yearbook, CD-ROM (Washington, D.C.: International Monetary Fund, July). Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi (2003), Governance Matters III: Governance Indicators for 1996 2002, World Bank Policy Research Working Paper no. 3106 (Washington, D.C.: World Bank, June). Krugman, Paul (1991), Geography and Trade (Leuven, Belgium, and Cambridge, Mass.: Leuven University Press and MIT University Press). Lewbel, Arthur (1997), Constructing Instruments for Regressions with Measurement Error When No Additional Data Are Available with an Application to Patents and R&D, Econometrica 65 (September): 120113. Lopez, R., and M. Schiff (1998), Migration and the Skill Composition of the Labour Force: The Impact of Trade Liberalization in LDCs, Canadian Journal of Economics 31 (May): 318 36. Ludema, R. D., and I. Wooton (1997), Regional Integration, Trade, and Migration: Are Demand Linkages Relevant in Europe? Centre for Econometric Policy Research Discussion Paper Series, no. 1656 (June). Mankiw, Gregory N., David Romer, and David N. Weil (1992), A Contribution to the Empirics of Economic Growth, The Quarterly Journal of Economics 107 (May): 407 37.

58

Thomas Osang

Martin, P. (2003), Economic Integration and Migration: The MexicoU.S. Case, World Institute for Development Economics Research Discussion Paper, no. 35 (April). North, Douglass C. (1993), Five Propositions about Institutional Change, Economic History Working Paper Archive no. 9309001 (St. Louis: EconWPA, September). (1994a), Institutions Matter, Economic History Working Paper Archive no. 9411004 (St. Louis: EconWPA, November). (1994b), Transaction Cost Through Time, Economic History Working Paper Archive no. 9411006 (St. Louis: EconWPA, November). (1994c), Institutional Change: A Framework for Analysis, Economic History Working Paper Archive no. 9412001 (St. Louis: EconWPA, December). Razin, A., and E. Sadka (1997), International Migration and International Trade, in Handbook of Population and Family Economics, ed. M. R. Rosenzweig and O. Stark (Amsterdam: Elsevier Science), 851 87. Robertson, R. (2005), Has NAFTA Increased Labor Market Integration Between the United States and Mexico? The World Bank Economic Review 19 (September): 425 48. Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi (2004), Institutions Rule: The Primacy of Institutions over Geography and Integration in Economic Development, Journal of Economic Growth 9 (June): 131 65. Russell, S. S., and M. S. Teitelbaum (1992), International Migration and International Trade, World Bank Discussion Paper no. 160 (Washington, D.C.: World Bank). Sachs, Jeffrey D. (2003), Institutions Dont Rule: Direct Effects of Geography on Per Capita Income, NBER Working Paper no. 9490 (Cambridge, Mass.: National Bureau of Economic Research, February). Sachs, Jeffrey D., and Andrew M. Warner (1995), Economic Reform and the Process of Global Integration, Brookings Papers on Economic Activity 1: 1 118. Smith, J. P., and B. Edmonston, ed. (1997), The New Americans: Economic, Demographic, and Fiscal Effects of Immigration (Washington, D.C.: National Academies Press, May).

External and Internal Determinants of Development

59

Stark, Oded, Christian Helmenstein, and Alexia Prskawetz (1998), Human Capital Depletion, Human Capital Formation, and Migration: A Blessing or a Curse? Economic Letters 60: 363 67. Taylor, A. M. (1995), Growth and Convergence in the AsiaPacific Region: On the Role of Openness, Trade, and Migration, NBER Working Paper no. 5276 (Cambridge, Mass.: National Bureau of Economic Research, September). United Nations (1994), U.N. Population Division, New York. World Bank (2000), Global Development Finance CD-ROM (Washington, D.C.: World Bank). (2003), World Development Indicators CD-ROM (Washington, D.C.: World Bank).